In this bonus episode of Between Two Vulns, we pick up right where we left off last month—only this time, we’re letting AI run the lab (uh-oh). We’re chatting about how AI co-scientists might spark real scientific breakthroughs, the security black holes in agent-based architectures, and whether humanity is truly ready for an AI that can do all our chores. Spoiler: we’re not.

Hunker down, hackers—AI benchmarks just broke another record, quantum computing chips are threatening your GPU’s existence, and rumor has it Arnold might make a comeback (we kid, we kid). In this episode, Dan and Marcello dissect how AI is evolving faster than we can say “prompt injection.” Then Ethan steps in with a killer breakdown of Model File Vulnerabilities (MFVs) on huntr—perfect for those of you itching to exploit the next big bounty target.

Kicking off 2025, Dan and Marcello cover the latest in AI security: 🔹 NVIDIA’s Project Digits: A $3K supercomputer for running LLMs locally and securely. 🔹 OpenAI GPT-03: The model beating human performance on Arc AGI. 🔹 AI agents: Will 2025 be the year they replace employees? 🔹 Prompt injection: Why it’s still the #1 LLM security risk.

2024’s been a wild ride y'all, and we’re breaking it all down in this Between Two Vulns finale: 💥 OpenAI drops a $200/month GPT Pro plan—worth it or a cash grab? 💥 Anthropic tries to standardize agents with their Model Context Protocol. 💥 Google’s quantum chip solves problems that would take septillion years (and maybe proves the multiverse is real). Then Ethan walks us through exploiting Keras Lambda vulnerabilities, and we close with shoutouts to the huntrs who crushed it this year. 🏆